| Ovis2.5 Technical Report | ✓ Link | 92.9 | Ovis2.5-9B | 2025-08-15 |
| DoCoG: Mask-based Multi-Type Grounded Chain-of-Thought for Document QA | | 92.7 | DoCoG (Ours) | 2026-09-08 |
| Twin-T & TwintVQA: A Reliable Structure-Detail Separating VLM and a Comprehensive Benchmark for Chart and Table Tasks | ✓ Link | 91.88 | Twin-T-7B | 2026-06-03 |
| MiniMax-01: Scaling Foundation Models with Lightning Attention | ✓ Link | 91.7 | MiniMax-VL-01 | 2025-01-14 |
| Vero: An Open RL Recipe for General Visual Reasoning | ✓ Link | 91.6 | Vero-Qwen3I-8B | 2026-04-06 |
| PerceptionDLM: Parallel Region Perception with Multimodal Diffusion Language Models | ✓ Link | 91.6 | PerceptionDLM-Base | 2026-06-17 |
| Introducing Command A Vision: Multimodal AI built for business | | 90.9 | Command A Vision | 2025-07-31 |
| Introducing Claude 3.5 Sonnet | | 90.8 | Claude 3.5 Sonnet | 2024-06-20 |
| iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning | ✓ Link | 90.4 | iVGR-Qwen3-VL-32B (ours) | 2026-05-29 |
| Qwen3-VL Technical Report | ✓ Link | 90.3 | Qwen3-VL-235B-A22B-Thinking | 2025-11-26 |
| Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence | ✓ Link | 90.3 | Nemotron 3 Nano Omni (Reasoning) | 2026-04-27 |
| The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation | ✓ Link | 90.0 | Llama 4 Maverick | 2025-04-05 |
| NVIDIA Nemotron Nano V2 VL | ✓ Link | 89.8 | Nemotron Nano V2 VL (Reasoning off) (extra data) | 2025-11-06 |
| InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models | ✓ Link | 89.7 | InternVL3-78B | 2025-04-14 |
| Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders | ✓ Link | 89.6 | Penguin-VL 8B | 2026-03-06 |
| Qwen2.5-VL Technical Report | ✓ Link | 89.5 | Qwen2.5-VL-72B | 2025-02-19 |
| Qwen3-Omni Technical Report | ✓ Link | 89.3 | Qwen3-Omni-Flash-Thinking | 2025-09-22 |
| Language-guided Frequency Modulation for Large Vision-Language Models | | 89.2 | LFM (w/ Qwen2.5-VL) | 2026-06-03 |
| SPOT-E: Test-Time Entropy Shaping with Visual Spotlights for Frozen VLMs | ✓ Link | 89.0 | SPOT-E (Qwen3-VL-8B) | 2026-06-18 |
| The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation | ✓ Link | 88.8 | Llama 4 Scout | 2025-04-05 |
| START: Spatial and Textual Learning for Chart Understanding | ✓ Link | 88.8 | START-RL | 2025-12-08 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 88.3 | Qwen2-VL-72B | 2024-09-18 |
| Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models | ✓ Link | 88.28 | SCOPED-8B (extra data) | 2026-10-15 |
| Qianfan-OCR: A Unified End-to-End Model for Document Intelligence | ✓ Link | 88.1 | Qianfan-OCR | 2026-03-11 |
| LongCat-Next: Lexicalizing Modalities as Discrete Tokens | ✓ Link | 88.0 | LongCat-Next (A3B, 68.5B) | 2026-03-29 |
| ERNIE 5.0 Technical Report | | 87.8 | ERNIE 5.0 | 2026-02-04 |
| MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning | ✓ Link | 87.6 | MMVistaReason-9B | 2026-10-22 |
| MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe | ✓ Link | 87.4 | MiniCPM-V 4.5 | 2025-09-16 |
| OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks | ✓ Link | 87.4 | OpenVLThinkerV2 (Ours) | 2026-04-09 |
| Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models | ✓ Link | 87.3 | Molmo 72B | 2024-09-25 |
| Qwen2.5-VL Technical Report | ✓ Link | 87.3 | Qwen2.5-VL-7B | 2025-02-19 |
| Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context | ✓ Link | 87.2 | Gemini 1.5 Pro | 2024-03-08 |
| ERNIE 4.5 Technical Report | ✓ Link | 86.8 | ERNIE-4.5-VL-424B-A47B | 2025-06-30 |
| Innovator-VL: A Multimodal Large Language Model for Scientific Discovery | ✓ Link | 86.8 | Innovator-VL (8B-Instruct) | 2026-01-27 |
| TACO: Tool-Augmented Credit Optimization for Agentic Tool Use | | 86.7 | TACO (Qwen2.5-VL-7B, 7B) | 2026-06-29 |
| ResAdapt: Adaptive Resolution for Efficient Multimodal Reasoning | ✓ Link | 86.6 | Qwen2.5-VL-7B + ResAdapt-RL (zero-shot, 105% retention) | 2026-03-30 |
| Phrase-Grounding-Aware Supervised Fine-Tuning for Chart Recognition via Side-Masked Attention | ✓ Link | 86.5 | PGA-SFT (Qwen2.5-VL-7B) | 2026-06-03 |
| Improving Vision-language Models with Perception-centric Process Reward Models | ✓ Link | 86.48 | Qwen2.5-VL (3B) with Perceval | 2026-04-27 |
| Aria: An Open Multimodal Native Mixture-of-Experts Model | ✓ Link | 86.4 | Aria | 2024-10-08 |
| Reinforced Attention Learning | | 86.4 | RAL | 2026-02-04 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 86.3 | VideoLLaMA3 7B | 2025-01-22 |
| Kwai Keye-VL Technical Report | ✓ Link | 86.3 | Keye-VL-8B | 2025-07-02 |
| FocusVTC: Efficient and High-Performance Visual Text Compression with Adaptive Resolution | ✓ Link | 86.28 | FocusVTC | 2026-09-29 |
| From Pixels to Words -- Towards Native One-Vision Models at Scale | ✓ Link | 86.2 | NEO-ov 9B | 2026-05-27 |
| Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation | ✓ Link | 86.2 | Intern Lumina U2 (NVIDIA) | 2026-10-14 |
| Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding | ✓ Link | 86.1 | Molmo2-4B | 2026-01-15 |
| Insight-V++: Towards Advanced Long-Chain Visual Reasoning with Multimodal Large Language Models | | 86.1 | Insight-V++ | 2026-03-18 |
| LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence | ✓ Link | 85.9 | LLaVA-OneVision-2-8B | 2026-05-25 |
| The Llama 3 Herd of Models | ✓ Link | 85.8 | Llama 3-V (405B) | 2024-07-31 |
| Hello GPT-4o | | 85.7 | GPT-4o | 2024-05-13 |
| Llama 3.2: Revolutionizing edge AI and vision with open, customizable models | ✓ Link | 85.5 | Llama 3.2 90B Vision Instruct | 2024-09-25 |
| PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding | ✓ Link | 85.5 | PLM-8B | 2025-04-17 |
| Beyond Correctness: Benchmarking and Aligning Response Behaviors in Hybrid-Thinking MLLMs | | 85.42 | Qwen3-VL-4B + PatternRL | 2026-08-17 |
| Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models | ✓ Link | 85.41 | VISE (Qwen3-VL-8B) | 2026-06-25 |
| Qwen2.5-Omni Technical Report | ✓ Link | 85.3 | Qwen2.5-Omni-7B | 2025-03-26 |
| FastVLM: Efficient Vision Encoding for Vision Language Models | ✓ Link | 85.0 | FastVLM-7B | 2024-12-17 |
| Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model | ✓ Link | 84.88 | Mage-VL-4B | 2026-07-27 |
| LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents | ✓ Link | 84.8 | LLaDA-UI (pre-training) | 2026-09-09 |
| Toward Cognitive Supersensing in Multimodal Large Language Model | ✓ Link | 84.7 | CogSense-8B | 2026-02-02 |
| Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone | ✓ Link | 84.5 | Dream-VL | 2025-12-27 |
| DiffusionVL: Translating Any Autoregressive Models into Diffusion Vision Language Models | ✓ Link | 84.2 | DiffusionVL | 2025-12-17 |
| M3Grounder: Mask-Based Multi-Span and Multi-Granular Grounding for Document QA | | 83.9 | M3Grounder-Q | 2026-06-03 |
| Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals? | ✓ Link | 83.9 | FVQ (8B) | 2026-10-27 |
| LLaVA-OneVision: Easy Visual Task Transfer | ✓ Link | 83.7 | LLaVA-OneVision-72B | 2024-08-06 |
| Phi-4-reasoning-vision-15B Technical Report | ✓ Link | 83.3 | Phi-4-reasoning-vision-15B | 2026-03-04 |
| Multimodal OCR: Parse Anything from Documents | ✓ Link | 83.2 | dots.mocr | 2026-03-13 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 83.0 | Qwen2-VL-7B | 2024-09-18 |
| Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models | ✓ Link | 83.0 | Embodied-R1.5 | 2026-06-09 |
| PixelUMM: Encoder-Free Unified Image and Video Understanding and Generation | ✓ Link | 82.96 | PixelUMM 8B MoT | 2026-09-29 |
| NaVLM-PVC: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs | | 82.8 | NaVLM-PVC | 2026-09-08 |
| Structure-Token Evidence-Anchored Reasoning for Scientific Chart Understanding | | 82.70 | STEER | 2026-09-07 |
| MODUS: Decoder-Only Any-to-Any Modeling of Diverse Modalities | ✓ Link | 82.45 | Modus | 2026-07-28 |
| ROSE: Rotate Your Large Language Model to See | | 82.4 | ROSE-7B (extra data) | 2026-06-03 |
| ZAYA1-VL-8B Technical Report | ✓ Link | 82.2 | ZAYA1-VL-8B | 2026-05-08 |
| From Pixels to Words -- Towards Native Vision-Language Primitives at Scale | ✓ Link | 82.1 | NEO-9B | 2025-10-16 |
| InfiniteVL: Synergizing Linear and Sparse Attention for Highly-Efficient, Unlimited-Input Vision-Language Models | ✓ Link | 82.0 | InfiniteVL-4B (extra data) | 2025-12-09 |
| Sparse-LaViDa: Sparse Multimodal Discrete Diffusion Language Models | | 82.0 | Sparse-LaViDa | 2025-12-16 |
| Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone | ✓ Link | 81.8 | Phi-3.5-Vision | 2024-04-22 |
| Pixtral 12B | ✓ Link | 81.8 | Pixtral 12B | 2024-10-09 |
| Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs | ✓ Link | 81.4 | Phi-4-Multimodal | 2025-02-24 |
| Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs | | 81.3 | ChartPaLI-5B + PaLM 2-S | 2024-03-19 |
| CASA: Cross-Attention via Self-Attention for Efficient Vision-Language Fusion | ✓ Link | 81.3 | CAt+v (Qwen2.5-VL, 3B) | 2025-12-22 |
| Representation Forcing for Bottleneck-Free Unified Multimodal Models | | 81.3 | Pixel+RF | 2026-05-29 |
| Grounded 3D-Aware Spatial Vision-Language Modeling | | 81.01 | GR3D-8B | 2026-05-28 |
| Gemini: A Family of Highly Capable Multimodal Models | ✓ Link | 80.8 | Gemini Ultra | 2023-12-19 |
| MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding | ✓ Link | 80.8 | MoE-ViE-H/14 + Llama 3.1 8B (42M alignment) | 2026-08-18 |
| Look Where It Matters: High-Resolution Crops Retrieval for Efficient VLMs | ✓ Link | 80.64 | AwaRes | 2026-03-14 |
| LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model | ✓ Link | 80.1 | LLaDA2.0-Uni (16B) | 2026-04-22 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 79.8 | VideoLLaMA3 2B | 2025-01-22 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 79.3 | DePlot+FlanPaLM+Codex (PoT Self-Consistency) | 2022-12-20 |
| LLaDA-o: An Effective and Length-Adaptive Omni Diffusion Model | ✓ Link | 79.3 | LLaDA-o | 2026-03-01 |
| ProLaViT: Learning Progressive Latent Visual Thoughts in Structured Latent Space | ✓ Link | 78.99 | ProLaViT (Full) | 2026-07-03 |
| GPT-4 Technical Report | ✓ Link | 78.5 | GPT-4 (vision, 4-shot CoT) | 2023-03-15 |
| Gemma 3 Technical Report | ✓ Link | 78.0 | Gemma 3 (27B, IT) (4-shot) | 2025-03-25 |
| Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs | | 77.3 | ChartPaLI-5B | 2024-03-19 |
| Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders | ✓ Link | 76.9 | Penguin-VL 2B | 2026-03-06 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 76.7 | DePlot+Codex (PoT Self-Consistency) | 2022-12-20 |
| ScreenAI: A Vision-Language Model for UI and Infographics Understanding | ✓ Link | 76.7 | ScreenAI 5B (4.62 B params, w/ OCR) | 2024-02-07 |
| InternVL-U: Democratizing Unified Multimodal Models for Understanding, Reasoning, Generation and Editing | ✓ Link | 76.6 | InternVL-U (4B) | 2026-03-10 |
| Mini-Monkey: Alleviating the Semantic Sawtooth Effect for Lightweight MLLMs via Complementary Image Pyramid | ✓ Link | 76.5 | Mini-Monkey (extra data) | 2024-08-04 |
| AdaptVision: Efficient Vision-Language Models via Adaptive Visual Acquisition | ✓ Link | 75.92 | AdaptVision (Qwen2.5-VL-7B-Instruct) | 2025-12-03 |
| Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification | ✓ Link | 75.9 | UniAR | 2026-06-16 |
| Cheers: Decoupling Patch Details from Semantic Representations Enables Unified Multimodal Comprehension and Generation | ✓ Link | 75.7 | Cheers (1.5B) | 2026-03-13 |
| Mobile-O: Unified Multimodal Understanding and Generation on Mobile Device | ✓ Link | 75.2 | Mobile-O-0.5B | 2026-02-23 |
| T5Gemma 2: Seeing, Reading, and Understanding Longer | | 75.0 | T5Gemma 2 4B-4B (post-trained) | 2025-12-16 |
| CoLT: Teaching Multi-Modal Models to Think with Chain of Latent Thoughts | ✓ Link | 74.7 | CoLT | 2026-06-30 |
| Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts | | 74.6 | SMoLA-PaLI-X Specialist Model | 2023-12-01 |
| LoopVL: Recurrent Visual Intelligence | ✓ Link | 74.52 | LoopVL-1B | 2026-09-29 |
| TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models | ✓ Link | 74.3 | Tuna | 2025-12-01 |
| Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts | | 73.8 | SMoLA-PaLI-X Generalist Model | 2023-12-01 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 73.5 | Qwen2-VL-2B | 2024-09-18 |
| Synthesize Step-by-Step: Tools Templates and LLMs as Data Generators for Reasoning-Based Chart VQA | | 72.64 | MatCha4096 + LaMenDa | 2024-01-01 |
| Non-Natural Image Understanding with Advancing Frequency-based Vision Encoders | | 72.5 | EDGE (extra data) | 2025-06-11 |
| PaLI-X: On Scaling up a Multilingual Vision and Language Model | ✓ Link | 72.3 | PaLI-X (Single-task FT w/ OCR) | 2023-05-29 |
| LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models | ✓ Link | 71.2 | LLaVA-NeXT-Interleave 14B | 2024-07-10 |
| PaLI-X: On Scaling up a Multilingual Vision and Language Model | ✓ Link | 70.9 | PaLI-X (Single-task FT) | 2023-05-29 |
| PaLI-X: On Scaling up a Multilingual Vision and Language Model | ✓ Link | 70.6 | PaLI-X (Multi-task FT) | 2023-05-29 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 70.5 | DePlot+FlanPaLM (Self-Consistency) | 2022-12-20 |
| mPLUG-DocOwl 1.5: Unified Structure Learning for OCR-free Document Understanding | ✓ Link | 70.5 | DocOwl 1.5 (extra data) | 2024-03-19 |
| PaLI-3 Vision Language Models: Smaller, Faster, Stronger | ✓ Link | 70 | PaLI-3 | 2023-10-13 |
| mPLUG-DocOwl2: High-resolution Compressing for OCR-free Multi-page Document Understanding | ✓ Link | 70.0 | DocOwl2 | 2024-09-05 |
| PaLI-3 Vision Language Models: Smaller, Faster, Stronger | ✓ Link | 69.5 | PaLI-3 (w/ OCR) | 2023-10-13 |
| SmolVLM: Redefining small and efficient multimodal models | ✓ Link | 68.7 | SmolVLM 2.2B | 2025-04-07 |
| Emu3: Next-Token Prediction is All You Need | ✓ Link | 68.6 | Emu3 (extra data) | 2024-09-27 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 67.3 | DePlot+FlanPaLM (CoT) | 2022-12-20 |
| Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond | ✓ Link | 66.3 | Qwen-VL-Chat | 2023-08-24 |
| UniChart: A Universal Vision-language Pretrained Model for Chart Comprehension and Reasoning | ✓ Link | 66.24 | UniChart | 2023-05-24 |
| Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond | ✓ Link | 65.7 | Qwen-VL | 2023-08-24 |
| StructChart: On the Schema, Metric, and Augmentation for Visual Chart Understanding | ✓ Link | 65.3 | StructChart+GPT3.5 (STR ChartQA+SimChart9K) | 2023-09-20 |
| VISion On Request: Enhanced VLLM efficiency with sparse, dynamically selected, vision-language interactions | | 65.2 | VISOR | 2026-03-24 |
| MatCha: Enhancing Visual Language Pretraining with Math Reasoning and Chart Derendering | ✓ Link | 64.2 | MatCha | 2022-12-19 |
| KOSMOS-2.5: A Multimodal Literate Model | ✓ Link | 62.3 | Kosmos-2.5-chat | 2023-09-20 |
| StructChart: On the Schema, Metric, and Augmentation for Visual Chart Understanding | ✓ Link | 60.7 | StructChart+GPT3.5 (STR) | 2023-09-20 |
| Preventing Expert Collapse in MoE-dVLMs via Modality-Wise Norm Alignment | | 59.0 | TowerAlign | 2026-09-08 |
| Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding | ✓ Link | 58.6 | Pix2Struct-large | 2022-10-07 |
| mPLUG-DocOwl: Modularized Multimodal Large Language Model for Document Understanding | ✓ Link | 57.4 | mPLUG-DocOwl (extra data) | 2023-07-04 |
| Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding | ✓ Link | 56.0 | Pix2Struct-base | 2022-10-07 |
| ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning | ✓ Link | 45.5 | VisionTapas-OCR | 2022-03-19 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 42.3 | DePlot+GPT3 (Self-Consistency) | 2022-12-20 |
| DePlot: One-shot visual language reasoning by plot-to-table translation | ✓ Link | 36.9 | DePlot+GPT3 (CoT) | 2022-12-20 |
| Rethinking Projector Training in Multimodal LLMs | ✓ Link | 17.9 | PORTAL (Vicuna 1.5 7B + SigLIP) | 2026-10-13 |